This report describes the clinical characteristics of psychotic patients who received a 6-month longitudinal research diagnosis of psychosis not otherwise specified (NOS) or for whom no consensus diagnosis was reached. The reasons why these subjects could not be classified into a specific DSM-III-R category, their classification under the proposed DSM-IV criteria, their reclassification at 24-month follow-up, and differences between these groups and patients with schizophrenia and affective disorders in demographic characteristics, initial clinical features, and short-term course are explored. Data were drawn from the first phase of the Suffolk County Mental Health Project. Longitudinal consensus procedures were used to derive 6-and 24-month DSM-III-R diagnoses based on information from a structured diagnostic interview, an interview with the patient's clinician, the medical record and discharge summary, and significant others. Thirteen subjects (4.7%) received a diagnosis of psychosis NOS, and 12 (4.3%) had no consensus diagnosis. Seven with psychosis NOS had an acute onset with rapid remission; this subgroup met DSM-IV criteria for brief psychosis without stressors. As a group, the psychosis NOS subjects were significantly older and had a lower rate of lifetime alcohol abuse/dependence than the schizophrenic and affective disorder groups. Their short-term course was significantly better than that of the schizophrenics and similar to that of patients with an affective disorder. Subjects with no consensus diagnosis were more likely to have lifetime drug abuse/dependence than the other two groups. Compared with patients with an affective disorder, these subjects showed poorer psychosocial functioning before and 6 months after hospitalization; these levels of functioning were not significantly different from those of schizophrenic subjects.